9 papers
BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation
Ning Li, Zixuan Guo, Yan Xu +7
Hallucinations remain a major obstacle to deploying large language models (LLMs) in knowledge-intensive settings, where generated responses must be faithfully grounded in provided…
Breaking the Tokenizer Barrier: On-Policy Distillation across Model Families
Yifan Niu, Han Xiao, Dongyi Liu +4
On-Policy Distillation (OPD) has become a core technique in the post-training of Large Language Models (LLMs) for transferring knowledge from domain experts to student models. Howe…
Efficient Scaling of LLM Training with Flexible Context Parallelism
Yifan Niu, Han Xiao, Dongyi Liu +2
Scaling long-context capabilities is crucial for Large Language Models (LLMs). However, real-world data contain a large number of sequences with heterogeneous lengths. Existing tra…
PiCA: Pivot-Based Credit Assignment for Search Agentic Reinforcement Learning
Dongyi Liu, Yifan Niu, Qinwen Wang +2
Large Language Model (LLM)-based search agents trained with reinforcement learning (RL) have significantly improved the performance of knowledge-intensive tasks. However, existing…
IBCircuit: Towards Holistic Circuit Discovery with Information Bottleneck
Tian Bian, Yifan Niu, Chaohao Yuan +7
Circuit discovery has recently attracted attention as a potential research direction to explain the non-trivial behaviors of language models. It aims to find the computational subg…
Mitigating the Safety Alignment Tax with Null-Space Constrained Policy Optimization
Yifan Niu, Han Xiao, Dongyi Liu +2
As Large Language Models (LLMs) are increasingly deployed in real-world applications, it is important to ensure their behaviors align with human values, societal norms, and ethical…